Recent content by buupq

  1. B

    Non-negative matrix factorization code

    Hello, I'm looking for the non-negative matrix factorization (NNMF) source code. I checked several linear algebra libraries (e.g., LaPack, mkl), but it seems that this subroutine is not available. Does anyone know where I can find this source code...
  2. B

    Create Row-Orthonormal Matrix | m > n

    Thanks very much StoneTemplePython! I'll try this.
  3. B

    Create Row-Orthonormal Matrix | m > n

    The approximation you gave is the output that I got from Singular value decomposition. What I try to approximate is the product ##A_{m\times n} A^T_{n\times m}## is as close as possible to the identity matrix ##I_m##, but there should be no columns/rows with only zero values. The criteria can...
  4. B

    Create Row-Orthonormal Matrix | m > n

    Yes, you are right. The ##A_{m\times n} A^T_{n\times m} = I_{m\times m}## does not exist if the number of rows > columns ##(m > n)## due to the linear dependence. Please see the rectangular matrix in the wiki page. https://en.wikipedia.org/wiki/Orthogonal_matrix. Yeah, my question is a bit...
  5. B

    Create Row-Orthonormal Matrix | m > n

    we have a set of eigenvalues, each correspondings to an eigenvector. What I mentioned was the collection of eigenvectors. If you use LAPACK or some mathlib, the output is usually a matrix of eigenvectors, matrix of eigenvalues, not just single one unless you request the output of only one.
  6. B

    Create Row-Orthonormal Matrix | m > n

    Hello, I'm looking for a way to create an approximate row-orthonormal matrix with the number of rows (m) > the number of columns (n); i.e., finding A(mxn) so that A(mxn) . A^T(nxm) = I(mxm). I used singular value decomposition (e.g., DGESVD in mkl mathlib), but what I actually got was an...
  7. B

    Meet Buu Pham: Rising Star in Computational Chemistry

    graduate student in computational chemistry.